Digitizing Historical Regesta: A Comparative Analysis of Computer Vision and Neural Network-Based OCR Methods

Regesta are brief abstracts of charters or other diplomatic documents. They constitute a vital resource for historians, offering a concise overview of primary legal documents and facilitating research. Digital processing enhances accessibility and expands research capabilities. This poster presents a comparative analysis of classical computer vision techniques and neural network-based methods for the automatic transcription of regesta collections. Despite the superior accuracy of neural networks, classical methods offer a more cost-effective alternative. This study uses Philipp Jaffé's "Regesta Pontificum Romanorum" from the "Forming Europe" project to evaluate this trade-off. The complexity of the tabular layout and text segmentation represent significant challenges in this example. In such cases, classical techniques may remain a viable option, offering practical alternatives for cost-effective transcription.